The physical characterization of exoplanets will require to take spectra at several orbital positions. For that purpose, a direct imaging capability is necessary. Direct imaging requires an efficient stellar suppression mechanism, associated with an ultrasmooth telescope. We show that before future large space missions (interferometer, 4-8 m class coronograph, external occulter or Fresnel imager), direct imaging of giant planets and close-by super-Earth are at the cross-road of a high scientific interest and a reasonable feasibility. The scientific interest lies in the fact that super-Earths share common geophysical attributes with Earths. They already begin to be detected by radial velocity (RV) and, together with giant planets, they have a larger area than Earths, making them detectable with a 1.5-2 m class telescope in reflected light. We propose such a (space) telescope be a first step before large direct imaging missions.